My AI chatbot kept failing until I saw what I actually wrote
Spent 3 weeks blaming the model for giving me junk replies, then I reread my prompt history and realized every single one was full of vague words like 'good' and 'make it better.' Has anyone else caught themselves writing prompts so sloppy that the AI basically had to guess what you meant?
Oh man, I did the exact same thing last year. I kept typing stuff like "write me a better version" and then getting mad when it came back almost the same. Once I started pasting the actual text I wanted changed and saying "cut this to 3 sentences and keep the part about the deadline" the replies got SO much better. The trick for me was pretending I was giving instructions to a brand new hire who knows nothing about the task. Say the exact thing you want, the length, and the tone, then it has nowhere to guess wrong. Sloppy in, sloppy out, every time.